FULL AUTOMATION OF QUANTITATION OF IN-VIVO NMR BY LPSVD(CR) AND EPLPSVD

A full automatic and reliable method for NMR quantification is propose d. It is based on the linear prediction singular-value decomposition w ith continuous regularization [LPSVD(CR)] and on the enhancement proce dure (EP) followed by LPSVD (EPLPSVD). Advantages of both methods have been considered. LPSVD(CR) has been used to find a good estimate for the number of spectral components. A significance test allows refining this number, which is then used in the EPLPSVD method. A Monte Carlo study has proved that EPLPSVD is a maximum-likelihood estimator. This new algorithm leads to reliable estimates for the spectral parameters even for signals with a low signal-to-noise ratio. No a priori knowled ge is needed. (C) 1994 Academic Press, Inc.